Note: This Research Bulletin is a source paper for RB 144 and will not be published in its present form.
In the first yearbook of Global Civil Society (Anheier et al. 2001a) the geography of its subject is given prominence. In the introductory chapter, the editors note that 'in particular, one of the most striking findings of the Yearbook is that global civil society is heavily concentrated in north-western Europe' (Anheier et al. 2001b: 7 (emphasis in the original)). They illustrate this with a table that identifies the top countries that are the 'focal points' of globalisation, international rule of law and global civil society (Anheier et al. 2001b: 8-9). Three indices of each category are included and the eleven countries that appear in six or more lists are highlighted: Austria, Denmark, Finland, Germany, Iceland, Luxembourg, Netherlands, Norway, Sweden, Switzerland, and the UK. The 'heavily concentrated' geography could hardly be more clear-cut. But is it? Is this the right 'space' for measuring these activities? In Castells' (1996) terminology, is this 'space of places' - countries - really how geographical focal points should be identified?
A preliminary answer to this question is provided by scrutinising Anheier et al.'s (2001b: 9) list of NGO host countries in which Belgium is ranked first in terms of NGO density per million of population.1 Why Belgium? It is not necessary to go back to the original data to realise that the process operating here is a desire of NGOs to be in and around Brussels, not because it is in Belgium but because it is the political and administrative centre of the EU. Clearly it is the city that is the attraction, not the country. In general, it can be noted that it is cities as coordinators of 'spaces of flows' (Castells 1996) rather than countries that constitute geographical focal points. In the second Global Civil Society yearbook (Glasius, Kaldor and Anheier 2002), Sassen (2002) makes the case for a city-centred interpretation of global civil society and her approach is developed in some detail in this paper. Specifically, the global location strategies of a large number of NGOS are measured and analysed within the framework of the contemporary world city network. The argument is that NGOs as major constituents of the global civil society organise their activities through cities across the world thus creating their particular space of flows through the world city network (Taylor 2004). In this way we place the 'north-western European concentration' in the context of a new global geography.
The argument is developed in seven steps. In the first part, studying NGOs is justified for understanding both global civil society and global governance and the role of NGOs as a link between these two spheres of global activity is briefly addressed. This is followed by a discussion of why geography - space and place - is seen as important for NGOs as a key element of global civil society to justify subsequent spatial measurement and analysis. From this definition of a space of flows, a model of the world city network is introduced in the third part. An interlocking network structure has been previously employed for world city network analyses; here it is used with NGOs in the role of 'interlockers' of cities. To operationalize the model requires specific data needs. In the fourth section the question of the dearth of global-level data when not studying countries is addressed and new data collection is described. This data is then analysed in two different ways. In section five, cities are measured in terms of their overall connectivity in the city network defined by NGOs. This provides a detailed look at the nodes in the NGO global space of flows and therefore allows for a more detailed geography of global civil society from that described through ranking countries. In the second analysis, in section six, the new model and data are used to go beyond identification of one particular geographical concentration. Ten separate organizational tendencies are identified thus suggesting that global civil society is more geographically dispersed - more global, in fact - than initially appears to be the case. There is a short conclusion in which two points are emphasized: first, the complexity of the new geography, and second, the limitations of this type of extensive research.
NGOs, GLOBAL GOVERNANCE AND GLOBAL CIVIL SOCIETY
One of the most illuminating ways of understanding the nature of contemporary globalization is to contrast it with what went before. This prior 'international order' is usually termed the 'Westphalia system' after the mid-seventeenth century treaties that confirmed the modern sovereign state as the prime player in international relations. In the basic version of this model, states compete with one another to create an international politics wherein the competition is ameliorated by governments adhering to the rules and norms of international institutions such as international law and, since the mid-twentieth century, the UN family of organizations. Below this 'level', within states, there are civil societies, conglomerates of social associations that make social life possible through individuals adhering to the rules and norms of 'society'. This simple 'two level' model of social organization evolved through three hundred years to dominate the twentieth century world order. Globalization challenges this international picture of the world by emphasizing transnational processes that are no respecters of political boundaries. In particular, contemporary global governance and global civil society transcend the traditional 'levels' of political and social activity. Although these two new concepts have very different provenances, they converge remarkably in their structures wherein networks replace territories.
According to Roseneau (1995: 182) 'Global governance is the sum of myriad - literally millions - control mechanisms driven by different histories, goals, structures, and processes'. This differs starkly from Westphalian international relations not only in its inherent complexity but also in the nature of the politics. As Roseneau tells it, traditional command and hierarchy are challenged by 'steering' behaviours in a framework of more horizontal relations, in short through network structures. States and their institutions, notably the UN, are part of this new complexity, but only part, for there are numerous other 'channels' of political behaviours. These are the many transnational organizations and movements that have mushroomed in recent years to create a worldwide informal politics of issues. And at the centre of the latter we find NGOs as the most powerful operators in this transnational politics. In terms of practice, therefore, it is the combination of these formal (state) and informal institutions as 'a set of interlocking but separate bodies' that 'produce the system of global governance' (Halliday 2000: )
Although it is commonly agreed that global civil society is a 'fuzzy concept' (Anheier et al. 2001b: 11; An-Na'im 2002: Chandhoke 2002)) with its 'organizational infrastructure' still in a 'state of flux' (Anheier and Themudo 2002: 191), nevertheless Keane's (2001: 23) description provides the essence of the subject: 'Global civil society is a vast, interconnected, and multilayered social space that comprises many hundreds of self-directing or non-governmental institutions and ways of life'. Through its 'cross border networks' global civil society is constituted of 'chains of interactions linking the local, regional and planetary orders' (p. 24), This new social world is constituted by 'networks, coalitions, partnerships and social movements' but once again, NGOs stand out as the key operators (Anheier and Themudo 2002). Thus, in terms of practice, civil global society looks very much like the informal half of global governance: both are composed of diffused networks that transcend sovereign territories and NGOs featuring prominently in both.
This is grist to the mill for Chandhoke's (2002: 35) 'three sector fallacy' which is the common assumption that politics, economy and society are separate and distinct realms of action. Extrapolated to the global level, this fallacy specifies a global civil society as the 'third sector' 'which can not only be distinguished from but which is an alternative to both the state-centric international order and the networks of global markets' (p. 36). Chandhoke shows that global civil society to be inherently 'contaminated' by both the other two sectors (most obviously through massive state and private fundings of NGOs). Her emphasis on the limits of global civil society tempers claims for its radical transformative role; she interprets it as being firmly embedded in a particular post-Cold war conjuncture dominated by rich countries. Here she alludes to Anheier et al.'s (2001) geographical concentration of global civil society as possible proof that it is part of the wider globalization process encompassing a far from radical contemporary international politics and economics.
This does not mean that we should abandon using the three sectors as organizing frames for our understanding of contemporary globalization but it does mean that we must always consider inter-sector relations. This is where NGOs are so interesting appearing as they do at the heart of global goverance and global civil society arguments. One perspective is to argue that NGOs constitute the key linking institutions between global goverance and global civil society, taking on parts of each but not wholly belonging to one sphere or the other (Taylor 1999: 92-3). Their role is analogous to political parties in the previous era. Neither part of government nor national civil society, parties operated as conduits between political elites and society. From the time when governments relied on parties for their election and therefore power, policies were proposed for society by parties to government, and also propagated to society by parties for government. NGOs have now found themselves in a similar ambiguous position, most noticeably in their links to UN activities (Otto 1996), and links to states through the distribution and organization of the development aid programmes from 'North to South' across the world.
For partly pedagogic reasons, in the remainder of this paper I will treat NGOs as part of global civil society. It is not just that they are more important to the latter than to global goverance with its large formal political apparatus, but it is a fact that NGOs have come to be more identified with global civil society - in Chandhoke's (2002: 38) words, 'NGOs play a large-than-life role in global civil society'. So much so that it is often found to be necessary to remind readers that global civil society is actually more than the activities of just NGOs (e.g. Anheier et al. 2001: 4; Chandkoke 2002: 38; Anheier and Themudo 2002: 191). While accepting the latter, NGOs remain the obvious starting point for describing the geography of global civil society and thus will feature as the institutions we measure and analyse below.
NGOs IN A GLOBAL SPACE OF FLOWS
This essay began with reference to the geography of concentration identified in the introduction to the first Global Civil Society yearbook; this geography appears several times in the arguments of subsequent authors (Glasius and Kaldor 2002: 4; Chandhoke 2002: 37; An-Na'im 2002: 57). Confirming the importance of geography, it is An-Na'im (2002: 56-7) who spells out why this should be. He argues that the physical location of interactions within global civil society are vital because this reflects power relations, in particular the ability to set policy agendas. Thus the fact that north western Europe is the most secular region in the world will likely produce a global civil society in which the world's major religions, that dominate all other regions, are not adequately or equitable recognised.
But this argument does not tell us what form of space constitutes the new geography of global civil society. If globalization is premised on large-scale transnational processes, as we argued above, then territorial frameworks - states - are hardly appropriate units for reporting (Anheier 2001). This is less the case with global governance in which IGOs (intergovernmental organizations) are important so that territorial thinking is imposed on policy making (Shafer and Murphy 1998). In NGO activities, sovereign territories have to be taken into account, but the essence of their operations in their freedom from such boundary restrictions. As we have seen, NGOs are creating a global civil society through a myriad of networks.
The appropriate framework for understanding NGOs has been provided by Castells (1996) with his concept of a space of flows. He argues that from the 1970s a new network society has emerged based upon the enabling technologies that have resulted from the merging of computing and communication industries. This has created a new type of space because social relations no longer depend upon spatial contiguity. With the new technologies social activities can be organized simultaneously across the world thus opening up a new space of flows. In network society, the latter dominate the old spaces of places (such as states) because it provides a flexibility in activity that can simply by-pass fixed assets in territories. Castells describes his space of flows at several levels starting with the infrastructural level (wires, satellites and electronic pulses). Most social activity occurs in Castells' second level of 'nodes and hubs' through which transnational social organization is constructed. Thus there are global networks of medical researchers sharing projects but located in separate nodes (e.g. Rochester, NY and Paris) but who communicate electronically with each other incessantly, while meeting regularly at international conferences (hubs).
The institutions that have taken most advantage of the infrastructure level are, of course, MNCs (multinational corporations) now often referred to as global corporations. They have developed worldwide networks of production and marketing that have revolutionized economic geography (Dicken 2003). However, Townsend (1999: 613) has argued that NGOs are 'quite as transnational as Exxon'. Leading NGO operatives are equally part of what van der Pijl (1998) has called a 'transnational class' (Townsend et al. 2002: 830). The result is that, 'as in the business sphere, fads sweep the world; microcredit is almost as placeless as just-in-time production' (Townsend 1999: 616). Thus although the circuits of flows are through different networks, NGOs and MNCs are both integral to the same overall global space of flows that defines contemporary globalization.
NGO ORGANIZATION AND CITIES: AN INTERLOCKING NETWORK MODEL
Castells (1996: 415) identifies Sassen's (1991) work on the global city as the 'most direct illustration' of the nodes and hubs social layer in the space of flows. Her research identifies the production and provision of advanced producer services (financial and business services) as the key practices producing global cities. These service firms, MNCs in their own right, provide specialist services to other corporations and governments across the world through their professional knowledge and creative practitioners in such complex areas as multi-market advertising and trans-jurisdictional law. Sassen focuses on just the major global cities but Castells emphasises the fact the network connections are not limited to just a few 'top cities'. Rather there is a world city network of global service centres that produce and distribute advanced services across the world (Taylor 2001).
Sassen (1994) has argued that cities constitute strategic places in the development of a new world geography. Recently she has transferred this argument from the global services market to global civil society (Sassen 2002). Thus the 'strategic cross border geography that bypasses national states' is part of 'the infrastructure of global civil society' (p. 217). This is because cities provide a 'thick enabling environment' (p. 217) through which transnational and sub-national activities can be brought together:
Thus she is able to argue that a 'focus on cities allows us to capture, not only the upper, but also the lower circuits of globalisation' (Sassen 2002: 219). I build upon this very powerful argument here.
Neither Castells nor Sassen, or indeed the world cities literature in general, have specified the exact nature of the worldwide networks that they frequently invoke (Taylor 2004). This has meant that it has not been possible to measure and analyse the networks empirically until quite recently. Specification of the world city network as an interlocking network has opened up new analytical possibilities (Taylor 2001; 2004). Interlocking networks have three levels instead of the usual two: as well as the network and nodal levels, there is a subnodal level. In the case of the world city network drawing on Sassen's early work, the subnodal is occupied by advanced service firms who 'interlock' cities through their large office networks in major cities throughout the world. Service firms need so many offices so that they can offer a seamless service to their clients thus preserving their brand image and, hopefully, satisfying their global clients. This network model can be transferred to a global civil society framework by replacing global service firms by global NGOs.
The NGO contribution to world city network formation can therefore be specified as follows. This portion of the world city network is constituted by m global NGOs with offices distributed across n world cities. The level of activity performed by NGO j in city i is zij which is called the activity value. The array of activity values defines a m x n activity value matrix, Z. A portion of the world city network is derived from this matrix using the plausible conjecture that the larger an NGO's activity value in a city, the greater the number of the NGO's flows of information, knowledge, instruction, ideas, strategies, plans, etc. will emanate from that city to other cities. This assumption allows for relations between cities to be defined.
The initial relation between each pair of cities is given by
xab,j = zaj . zbj (1)
which describes relations between cities a and b in terms of NGO j. This defines an elemental interlock link between two cities. It is multiplicative because the potential quantity of flows between two cities rises geometrically with the quantity of activity provided in each city. From this the aggregate city interlock link between two cities can be derived as
xab = Sxab,j (2)
For each city there are n-1 such links, one to every other city. These can be used to define the interlock connectivity of a city so that
Ya = Sxai where a ¹ i (3)
This measure of connectivity picks up two features of a city's activity values. First, and most obviously, cities where NGOs locate offices with higher activity values are more connected. Second, and more subtly, if those high activity offices are for NGOs with very larger office networks then the city appears more connected. In other words a city with several large offices of NGOs that themselves have small networks will not be that well connected as measured by its interlock connectivity.
Interlock connectivity indicates the importance of a city in the world city network. In the discussion that follows empirical measures of interlock connectivity will be referred to as NGO network connectivity.
OPERATIONALIZING THE MODEL: DATA COLLECTION
For Townsend (1999: 617) the NGO community is a 'puzzle' because despite much research on specific NGOs, projects and themes, 'one dimension is missing': 'global maps of flows' within the NGO community. The reason for this research lacuna is spelt out clearly by Anheier (2001). He identifies the culprit as 'methodological nationalism':
This geographical co-incidence assumption is the embedded statism endemic to macro-social science that is reliant on 'state-istics' (Taylor 2003). According to Anheier (2001: 222) it has led to NGOs being 'treated as domestic agents for measurement purposes, therefore losing an essential aspect of their very raison d'etre.' However, in his quest to measure global civil society using official data Anheier (2001: 223) is forced to concede that 'the country becomes the de facto unit of analysis'. Hence the data in the first Global Civil Society yearbook (Anheier, Glasius and Kaldor 2001) are overwhelmingly about states, which is both unsatisfactory and very disappointing.
Obviously the answer to this problem is not to rely on official statistics. A beginning is made along this alternative route in the second Global Civil Society yearbook (Glasius, Kaldor and Anheier 2002). The new data comes in two forms. First, a ranking of cities in terms of numbers of NGOs present is provided by Glasius and Kaldor (2002: 6) which, they point out (p. 4), again emphasises the dominance of Europe (Brussels is ranked first and there are 12 further European cities in the top 20). This aggregate data is interesting but does not inform us about NGO networks. Second, networks are to be found in the yearbook's data sets where information about states is augmented by four maps of 'global networks'. However, here we are back to specific cases, interesting but providing no general guide to NGO networks across the world. In the analyses presented below these two forms of data are combined to create aggregate measurements of networks. This is made possible by the precise specification of the network model above. A key advantage of the specification is that it directs data collection to enable empirical description and analysis of the new geography.
The data requirement is to construct a NGOs x cities matrix, Z, as an array of 'service values'.2 This requires gathering information on the office networks of a large number of NGOs, recording their city locations, and assessing the importance of each office to create 'activity values'. This data collection exercise was conducted as follows:
The end result is a matrix defined by 74 NGOs and 178 cities listed in appendices A and B respectively. The resulting array of 'activity values' range from 0 (no presence of an NGO in a city) to 4, where each cell indicates the importance of a given city to the office network of a given NGO. Hence, every column describes a simple coding of an NGO's office location strategy, and every row describes a simple coding of the mix of NGO offices in a particular city. This is the matrix Z in the specification of an interlocking network model operationalized for empirical analyses.
CITY CONNECTIVITIES IN GLOBAL CIVIL SOCIETY
The most elementary way of defining the geography of global civil society with the new data is to look at the distribution of the 178 cities across major world regions. The first column in Table 1 lists these frequencies across 10 regions. This simple exercise provides the first evidence that the European concentration, so emphasized in previous discussions of the geography of global civil society, is not replicated in this study. Sub-Saharan Africa is the leading region in terms of cities featured in our data, followed by Latin America. Western Europe does appear in third place and certainly is far ahead of northern America in featured cities, which is often the contrast made when referring to the European concentration (e.g. Ahheier et al. 2001b: 7). But there can be no doubt that the data collection methodology employed here has shown that cities in the erstwhile 'third world' dominate the selection. This is partly explained by the fact that NGO activity is state-related and therefore their office networks prominently feature capital cities. As the third and fifth columns of Table 1 show, 139 (78%) of the cities featured are state capitals. Since the majority of the world's states are from the erstwhile third world, then the pattern of regional frequencies among our 178 cities could be merely a consequence of this attraction to capital cities. Certainly the dearth of northern American cities in our list may result from there being only two capital cities in the region. This would also account for more cities featured from Western Europe. However, notice that these two richest regions do feature most non-capital cities; this may indicate a particular importance of these two regions for NGOs. To take this discussion further it is necessary to consider NGO network connectivities.
Using equation (3), NGO network connectivities have been computed for all 178 cities. This is a network measure that assesses each city in terms of the density of its links to other cities through NGOs. Thus it provides a first indication of a space of flows in the geography of global civil society. It could be, for instance, that the prevalence of erstwhile third world cities in our list includes large numbers with very low connectivities. In other words there might still be a concentration of European cities among the highly connected cities. This is tested in columns 2, 4 and 6 of Table 1, where the top 50 cities in terms of NGO network connectivity are distributed across the world regions. Note that the ranking of the top two regions remain the same, and Western Europe is now only equal third with Pacific Asia. This shows reasonably conclusively that the NGO portion of the world city network is relatively biased towards poorer parts of the world.
The details of the upper echelon of NGO connectivities are illustrated in Table 2, where the top 25 cities are the focus. It provides the second surprise of this analysis: this must be the first ever list of world cities headed by Nairobi! Simon (1995) has previously shown Nairobi to be important as a world city in African and third world comparisons but not at a global level. Nairobi does appear in Glasius and Kaldor's (2002: 6) list of cities hosting NGOs but only in 16th position. However we should be wary of making direct comparisons with their listing for two main reasons. First, the universe of NGOs is different: their 'international and internationally-orientated NGOs' is not as restricted as the 'global NGOs' selected above. Second, their ranking only counts the number of NGOs in a city and takes no account of their office networks beyond a particular city. Thus what the results in Table 2 indicate are cities housing NGOs with very large office networks (i.e. the important 'interlockers' in our network model). Hence, although Brussels remains important, ranked second in the new ordering, European cities in general fair poorly: from 13 in Glasius and Kaldor's top 20 to only 6 out of 25 in Table 2. This suggests that while there may be a surfeit of NGOs in Europe, many have few inter-city links so that this does not translate into a surfeit of well-connected cities in global civil society.
Table 2 also reinforces the importance of capital cities to NGOs: there are only two non-capital cities in the list. However both of these cities, New York and Geneva, have special political roles through the UN organization. Thus, this reinforces the finding that there is a strong propensity of NGOs to be attracted to loci of political power. It reminds us that NGOs are important as agents in global goverance as well as creators of global civil society.
One big advantage of careful specification and measurement in a network analysis is that the NGO results can be compared to other analyses of the world city network. In particular, the office networks of 100 global service firms have been coded over 315 cities and city connectivities computed (Taylor et al. et al. 2002a). These are termed global network connectivites because they represent the prime processes in world city network formation (Sassen 2001; Taylor 2004). NGO and global network connectivities are compared in Table 2. The differences really hammer home the distinctiveness of our results: only 7 of the cities in Table 2 appear in the top 25 for global network connectivity - Brussels, London, New York, Mexico City, Jakarta, Tokyo, and Buenos Aires. Many very important world cities are relatively unimportant for NGO connectivity: for example Paris is only ranked 36th, Amsterdam 91st, Frankfurt 139th, Los Angeles comes in at 162nd and Chicago does not make the 178. In total contrast, many cities that are important for NGO connectivity are quite minor world cities in economic terms: there are 6 differences of ranking of over 100 in Table 2. In all cases it reflects the particular importance of erstwhile third world cities in the NGO portion of the world city network.
The theoretical importance of these findings is very clear. As we have shown previously, the emphasis on a European concentration is part of an argument that is concerned for global civil society having a very similar geographical bias to the global economy and global geopolitics. In other words, it has been suggested that global civil society is little more than just another projection of the power of rich countries. Framed in a network analysis, where power is more diffuse and practiced in more subtle forms (Taylor 2002b), this disquiet for the distinctiveness of global civil society seems to be unwarranted. NGO network connectivities define a new and distinctive geography.
GEOGRAPHICAL CONFIGURATION OF GLOBAL CIVIL SOCIETY
Measuring connectivities indicates the importance of cities in the NGO portion of the world city network but it does not inform us about the patterns of relations between the cities. We need to answer Anheier's (2001: 227) question 'what kinds of networks do they (i.e. NGOs) form?' For instance, we know that Nairobi has a denser set of linkages than, say, London, but this does not indicate the actual roles that each city plays in the network, how its mix of NGO offices contribute to the overall functioning of the network. Such an investigation might well suggest that, contrary to the previous discussion, 'first world' cities have patterns of connections that suggest they play particularly important roles in the network above and beyond their actual NGO connectivity level. Such ideas have been explored for advanced producer services by applying a principal components analysis to the services/cities matrix (Taylor et al 2004; Taylor 2004). Here we use the same technique on the NGOs/cities matrix.
Principal components analysis is a means of data reduction; it treats a large number of variables and reduces them to a small number of 'composite variables' or components. The operation of the technique can be best envisaged geometrically. The input to the analysis is a data matrix consisting of p variables measured across k cases or objects. This matrix is treated as a p-dimensional space to which each variable contributes an axis. Each axis is one unit in length; this represents the spread of values (variance) of each variable so that the total variance is p, the same as the number of variables in the matrix. By analysing the co-variance (correlation) amongst the variables, an alternative set of axes of different lengths can be produced ranked by size. These are principal components that reorder the data as follows: the first (largest) component describes the biggest cluster of co-variance amongst the original variables, the second component the next biggest of cluster of co-variance, and so on down to a very small final pth component. The basic idea is that most of the variation in the data is revealed in the first few large components so that many small components can be discarded as unimportant. The result, therefore, is to transform an original p-dimensional variable space into a much smaller q-dimensional component space defined by selection of q principal components. The result is a parsimonious solution, converting a large number of variables p into a relatively small number of relevant components q that between them account for a relatively large proportion of the original variance. For example, in the analysis of producer service firms, the 100 variables (firm's location strategies) were reduced to just 6 principal components (composite location strategies) that accounted for 52% of the original variation (Taylor et al 2004; Taylor 2004). In other words, instead of describing 100 different patterns of service provision across cities, we were able to describe just six composite patterns that cover more than half the variability in the data.3
The basic results of a principal components analysis consist of three sets of new information.
Armed with this information, a researcher can generate a parsimonious description of a large data set through identification, description and interpretation of common patterns of variation.
In the analysis reported here the variables are NGO locational strategies recorded as a string of values between 0 and 4 across cities. The latter are the objects. The initial activity values matrix of 74 NGOs x 178 cities had to be reduced because of sparsity (excessive zeros) in the matrix - principal components analysis results are sensitive to very sparse matrices. To reduce data sparsity, the number of cities was reduced to 100 using total activity value scores to rank them: they are identified in appendix B where they are also arrayed as a cartogram (Figure B1) that is used to present findings below. NGOs were then eliminated if they had offices in less than 4 of the 100 cities. 58 NGOs survived this culling: they are identified in appendix A. Hence the analysis is based upon a reduced 58 (NGOs) x 100 (cities) data matrix. Although not including all the detailed data collected, the matrix does nevertheless contain 5,800 pieces of information (activity values), which is more than adequate to provide a reasonable indication of the geographical patterns in global NGO office networks.
In selecting the number of components to focus upon we considered both eigenvalues and the distribution of loadings.4 For the latter only loadings above 0.4 were taken as indicating a variable (NGO) making a significant contribution to a component. Using this loading threshold, components were only considered that had at least 3 variables contributing to a component's construction. Ten such components were thus identified and all had eigenvalues above 1.75, thus all contributing to at least 3% of the original variation. Between them the 10 components accounted for 51% of the variance in the data (i.e. we are describing just over half the variation in the data). All the NGOs with loadings above 0.4 for the 10 components are listed in the appendix C. Usually loadings are central to the identification of components and are, therefore, described in detail. Here, because our main concern is to depict new geographies, we will focus on the cities and their component scores, and only bring in the component loadings to supplement the interpretation and discussion.
The principal components analysis has created ten new 'composite variables' each of which defines an alternative office locational strategy. These are mapped as component scores for the cities as set out on the cartogram (Figure B1). As previously noted, since the scores are computed as standardised variables they are not constrained between 0 and 4 as the original variables were. In fact, for every component, approximately half the cities will have negative scores. This simply means that the negatively scoring cities have below average activity for that component. We focus on the positively scoring cities for each component which we interpret as defining ten global NGO fields of activity. Within each field cities are categorised into 4 strata.
The latter category is used to distinguish between two types of field of activity: those with articulator cities are vertical activity fields where the common locational strategy appears to be very hierarchical, and those without articulator cities are horizontal activity fields that appear to exhibit far less hierarchical tendencies in the common locational strategies. There are just 7 articulator cities that emerge from the analysis (Table 3) thus the vertical to horizontal ratio is 7 to 3.
Finally, the eigenvalues indicate the relative importance of the components in terms of the amount of common variance they encompass. These figures are provided in appendix C. We use them to define three levels of field of activity: components I and II are alpha fields of activity (with eigenvalues of 5.65 and 4.23 they account for 9.7% and 7.3% of the total variance respectively); components III and IV are beta fields of activity (with eigenvalues of 3.75 and 3.09 they account for 6.5% and 5.3% of the total variance respectively); and the remaining 6 components are designated gamma fields of activity (with eigenvalues ranging from 2.54 to 1.77 and variance accounted for ranging from 4.4% to 3%). These three levels are used to order the discussion of the 10 fields of activity.
Alpha Fields of Activity
London's vertical field of activity (component I) is the most vertical of all activity fields. Not only is London's component score the highest in the whole analysis (Table 3), the map of the field in Figure 1 shows that there are no major activity centres. This is the only component that has no cities in this strata. The other features of this field are (i) the absence of any other western European cities; (ii) the presence of three other major world cities, Los Angeles, Moscow and Tokyo; (iii) the dominance of poorer regions of the world but without a 'Commonwealth bias' i.e. south Asian cities feature prominently but Latin American cities are also well represented. Finally, it should be noted that most of the NGOs with high loadings on this component are, not surprisingly, those with London HQs (appendix C). Thus we can say that this is a 'North-South interactive' field where the former is embodied dominantly by London at the head of a common, extreme, hierarchical location strategy.
The horizontal field of UN associated activity (component II) is easily identified through the NGOs that load high (appendix C) where UN NGOs feature prominently. Clearly this UN association generates a horizontal field of activity with no articulator city but three predictable major activity centres: New York, Geneva and Cairo (Figure 2). The field itself is widespread with 18 standard centres (twice as many as component I) and just two conspicuous absentees: (i) London and other 'old' Commonwealth cities, and (ii) former 'COMECON' cities (only Moscow features as a minor centre). Thus we can say that this is another 'North-South interactive' field but without so much of a hierarchical tendency.
The contrast between the two maps of alpha fields of activity could hardly be starker. They represent two opposite types of NGO locational strategies across the old 'North-South divide'.
Beta Fields of Activity
Washington's vertical field of activity (component III) is the second most vertical field in the whole analysis: Washington's component score is the only one to be close to London's (Table 3) and there is just one major centre in the field (Figure 3). But this is where the similarity ends. The loadings show several NGOs with Washington HQs as would be expected: the NGOs in this articulator city can be interpreted as those promoting, and mopping up after, the 'Washington consensus' policy programmes. The geography being articulated is quite specific compared to other activity fields. Brussels is the major activity centre but it is former-COMECON cities that stand out in this field. East Asia and Latin America are also featured in the field but Africa is under-represented and south Asia and the Middle East do not feature at all. Not so much a North-South interaction, this is more like an East-West interactive activity field.
Geneva's vertical field of activity (component IV) features the third highest component score and, with only one major activity centre, this activity field depicts an office locational strategy approximately as vertical as Washington's field (Figure 4). The loadings show a quite different set of NGOS to the UN associated horizontal field, largely featuring important lobbying NGOs (appendix C). The pattern of cities in the field is very distinctive with Pacific Asia (including major centre Tokyo) and Australasia, and Latin America prominent while Africa and the Middle East are greatly under represented. This is a strange north-south interactive activity sans Africa.
Although these two beta fields are clearly distinctive, they do have a similar propensity: a specific avoidance of cities in the poorest regions of the world.
Gamma Fields of Activity
The horizontal field of Northern-focussed activity (component V) is organised through six major activity centre (Figure 5), three in Europe (Copenhagen, Paris, Rome), one in the Americas (Toronto), one in Pacific Asia (Hong Kong), and one in Africa (Nairobi). Note that only the latter is in the 'South'; this 'northern bias' is confirmed by the standard centres which are heavily concentrated in Europe. In fact there are only 4 standard offices in erstwhile 'third world' regions. If this is an office locational strategy encompassing a north-south interaction it is clearly very narrowly focussed within the latter zone.
The Western European/West African horizontal field of activity (component VI) is organised through just two major centres, Amsterdam and Rome (Figure 6). Although there is some notable representation in Latin America, the pattern is dominated by Western European and West African cities. This is the first office locational strategy that appears to link two specific regions in a north-south interaction.
Nairobi's vertical field of activity (component VII) is the first field articulated through a city outside Europe or North America (Figure 7). This confirms Nairobi's role as a leading NGO city: beyond its large number of inter-city connections (Table 1), it also operates with an important organisational role. However, this does not translate into an African-orientated field of activity. In fact quite the opposite emerges: Pacific Asia is well represented (the only two major centres in the field are Bangkok and Tokyo), the Americas are reasonably represented with Europe and Africa relatively under-represented. This is in some ways the obverse of the previous pattern. Thus this locational office strategy has Nairobi articulating African cities almost single-handedly, it is most definitely does show the city as part of an emerging African city network.
Brussels' vertical field of activity (component VIII) is another 'Northern' pattern with Geneva, Los Angeles and Rome as the only three major centres (Figure 8). There is an Eastern European concentration of standard centres but otherwise these are scattered across the world: as a north-south interaction this office locational strategy appears to favour a 'one city per region' distribution: Lome for Africa, Santiago for Latin America, Cairo for the Middle East, Kuala Lumpur for South East Asia.
Manila's vertical field of activity (component IX) is the second field articulated by a city outside Europe and North America (Figure 9). Unlike the Nairobi articulation, Manila's does include its own region (Pacific Asia) plus a strong showing in neighbouring South Asia. This office locational strategy exhibits an alternative 'one city per region' distribution in its standard offices beyond its concentration: Lagos for West Africa, Kampala for East Africa, Johannesburg for South Africa, Santiago for South America, and Mexico City for Central America.
Ottawa's vertical field of activity (component X) is remarkable for the fact of its articulator city. This is a city that ranks a lowly 144th in NGO network connectivity5 and yet appears in this role, albeit in the smallest component we consider. This shows conclusively how a reordering of the variance through a principal components analysis does pinpoint key organizational nodes in the NGO portion of the world city network beyond measurement of connectivity. The pattern of cities (Figure 10) features an Americas bias (two of the four major centres are here: Toronto and Santiago) with a smaller concentration in South East Asia. Europe is equally conspicuous due to its relatively low representation. This office locational strategy shows is one that largely frames a north-south interaction in a specific regional context.
The gamma fields of activity are most interesting for their identification of two erstwhile 'third world' cities as articulators. This can be down played by noting that the components are among the least important (appendix C), and that Manila and Nairobi, in particular, do not have high loadings amongst the articulator cities (Table2). However, I prefer to emphasize the opposite interpretation: it is very relevant that two cities from poorer countries in the world do achieve the status of articular cities in a potion of the world city network. This is unique to this type of global urban analysis with all articulators for both services and media having been found to be important world cities in the North (Kratke and Taylor 2003; Taylor et al. 2004).
There are two main conclusions that can be drawn from the research reported here, one empirical and one methodological.
The discussion began with the observation of a North West European concentration in global civil society and this simple geography has been challenged and found to be wanting by the results reported above. Focussing on NGOs and using a city-centred approach, we have produced measures of NGO city connectivities that indicate a worldwide network of cities underpinning the development of a global civil society. Some North West European cities do feature prominently, notably Brussels, London and Geneva, but the region does not dominate at all in the way that state-centred data had shown. This does not mean that there is no 'domination' of North over South that the original argument incorporating the European bias implied. The connectivity results must be moderated by the identifications of vertical fields of North-South interaction primarily articulated by Northern cities, and with the more horizontal fields of activity featuring largely Northern cities as major centres. But this is only part of the story. Whatever else the current rise of global processes may represent, they are networked and this means that power is to some extent diffused across the world. Thus it should come as no surprise that non-Northern cities are to be found as both articulator cities and major centres. It follows that a simple geography emphasizing just one region of the world was never going to properly capture what is going on its this sphere of global activity. The key point that our empirical results emphasize is that the geography of global civil society is very complex even when dealing with just one aspect of it in NGO organization.
This is an important finding based upon a set of unique results for studies of both global civil society and world cities. But we have to be very modest in our interpretations of how far we have gone along the road to understanding the geography of global civil society and the role of cities in its development. Our urban global analysis provides a fascinating glimpse of a complex geography but the results remain tantalizing, not in any sense definitive. This is because a quantitative, extensive research methodology can never uncover the way in which processes operate. The geography we have described provides increased understanding of the framework through which the processes of global civil society are operating and the outcomes of those operations. But we are no further advanced in our understanding of what the relations between cities mean in terms of the constitution of a global civil society. For instance, Townsend et al. (2002: 833), following Tvedt (1998), argue that flows within NGO organizations are a one way 'transmission belt' whose basic task is to feed 'management information' up the hierarchy. In city terms, this would mean that erstwhile 'third world cities' are merely transmission centres for information to be processed elsewhere. In other words, they remain 'static cities' in Jacobs (1984) terms, with dynamic cities in the NGO portion of the world city network remaining ensconced in the North. This implies the positive extensive findings reported above - for instance, Nairobi and Manila as articulator cities - is misleading at best. But like the initial simple regional geography, this simple network geography needs thorough testing to see whether it is hiding a much more complex set of relations. It most probably does but we will only know when more intensive research is carried out at a global scale to complement the extensive research reported here.
The data upon which this paper is based were collected by Troy Gravitt as part of his postgraduate internship at the Metropolitan Institute at Virginia Tech. Thanks are due to Troy for doing such a sound job and to the College of Architecture and Urban Studies at Virginia Tech for funding the internship.
Anheier, H. (2002). 'Measuring global civil society', Introducing global civil society', in H. Anheier, M. Glasius, and M. Kaldor (eds) Global Civil Society 2001. Oxford: Oxford University Press
Anheier, H., Glasius, M. and Kaldor, M. (eds) (2001a). Global Civil Society 2001. Oxford: Oxford University Press
Anheier, H., Glasius, M. and Kaldor, M. (eds) (2001b). 'Introducing global civil society', in H. Anheier, M. Glasius, and M. Kaldor (eds) Global Civil Society 2001. Oxford: Oxford University Press
Anheier, H. and Themudo, N. (2002) 'Organizational forms of global civil society: implications of going global', in M. Glasius, M. Kaldor, and H. Anheier (eds) Global Civil Society 2002. Oxford: Oxford University Press
An-Na'im, A. (2002). 'Religion and global civil society: inherent incompatibility or synergy and interdependence?', in M. Glasius, M. Kaldor, and H. Anheier (eds) Global Civil Society 2002. Oxford: Oxford University Press
Castells, M. (1996). The Rise of Network Society. Oxford: Blackwell
Chandhoke, N. (2002). 'The limits of global civil society', in M. Glasius, M. Kaldor, and H. Anheier (eds) Global Civil Society 2002. Oxford: Oxford University Press
Glasius, M. and Kaldor, M. (2002). 'The state of global civil society before and after September 11th', in M. Glasius, M. Kaldor, and H. Anheier (eds) Global Civil Society 2002. Oxford: Oxford University Press
Glasius, M., Kaldor, M. and Anheier, H. (eds) (2002). Global Civil Society 2002. Oxford: Oxford University Press
Halliday, F. (2000) 'Global governance: prospects and problems'. Citizenship Studies 4:
Kamat, S. (2003). 'The NGO phenomenon and political culture in the Third World'. Development 46: 88-93
Keane, J. (2002) 'Global civil society?', Introducing global civil society', in H. Anheier, M. Glasius, and M. Kaldor (eds) Global Civil Society 2001. Oxford: Oxford University Press
Kratke, S. and Taylor, P. J. (2003). 'A world geography of global media cities'. European Planning Studies
Otto, D. (1996) 'NGOs in the UN system: the emerging role of international civil society'. Human Rights Quarterly 18: 107-41
Pijl, van der K. (1998) Transnational Classes and International Relations. London: Routledge
Roseneau, J. N. (1995) 'Governance and democracy in a globalizing world', in D. Archibugi, D. Held and M. Kohler (eds) Re-imagining Political Community: Studies in Cosmopolitan Democracy. Cambridge: polity
Sassen, S. (1991). The Global City. Princeton, NJ: Princeton University Press
Sassen, S. (1994). Cities in a World Economy. Thousand Oaks, CA: Pine Forge
Sassen, S. (2002). 'Global cities and diasporic networks: microsites in global civil society', in M. Glasius, M. Kaldor, and H. Anheier (eds) Global Civil Society 2002. Oxford: Oxford University Press
Simon, D. (1995). 'The world city hypothesis: reflections from the periphery', in P. L. Knox and P. J. Taylor (eds) World Cities in a World-Economy. Cambridge: Cambridge University Press
Townsend, J. (1999). 'Are non-governmental organizations working in development a transnational community?'. Journal of International Development 11: 613-23
Townsend, J., Porter, G. and Mawdsley, E. (2002). 'The role of the transnational community of NGOs: governance or poverty reduction?' Journal of International Development 14: 829-39
Tvedt, T. (1998). Angels of Mercy or Devlopment Diplomats? NGOs and Foreign Aid. Oxford: James Curry
Taylor, P. J. (2001). 'Specification of the world city network'. Geographical Analysis 33: 181-94
Taylor, P. J. (2003) 'Metageographical moments: a geohistorical interpretation of embedded statism and globalization', in M A Tetreault, R A Denemark, K P Thomas and K Burch (eds.) Rethinking Political Global Economy: Emerging Issues, Unfolding Odysseys. London: Routledge
Taylor, P. J. (2004). World City Network: a Global Urban Analysis. London: Routledge
Taylor, P. J., Catalano, G., and Walker, D. R. F. (2002a). 'Measurement of the world city network'. Urban Studies 39: 2367-76
Taylor, P. J., Catalano, G., Walker, D. R. F. and Hoyler, M. (2002b). 'Diversity and power in the world city network'. Cities 19: 231-41
Taylor, P. J., Catalano, G., and Walker, D. R. F. (2004). 'Multiple globalisations: regional, hierarchical and sectoral articulations of global business services through world cities'. The Services Industries Journal
1. The NGOs they include are all international in scope (INGOs). The discussion of NGOs in this paper focuses upon their transnational activities and therefore the common synonym INGO is not used. For the empirical description and analysis in the paper, a category of global NGOs is defined.
2. The methodology employed has been previously used for studying advanced producer service firms (Taylor et al. 2002) and global media conglomerates (Kratke and Taylor 2003).
3. Note that principal components analysis is used as a statistical technique so that not all the variation in the data is accounted for. The variation that is unaccounted for is treated as uninteresting because it consists of patterns representing merely particular variables, rather than patterns common to several variables as described by the large selected principal components.
4. The standard procedure of applying a varimax rotation to initial solutions was employed to produce distinctive components.
5. Note that this city qualifies for the reduced data set because it is in the top 100 cities as measured by total activity value.
Table 1: Regional and political distribution of 178 NGO cities
*Regions are defined as follows:
Australasia: Australia, New Zealand and Pacific Island states
Central Asia: former USSR Asian republics plus Mongolia
Eastern Europe: former COMECON states in Europe plus former Yugoslavian states, Albania and Turkey
Latin America: from Mexico to Chile plus Caribbean states
North Africa/ western Asia: from Morocco to Iran
Northern America: Canada and USA
Pacific Asia: from Thailand to Japan
South Asia: from Pakistan to Burma
Sub-Saharan Africa: all African states without a Mediterranean coast
Western Europe: EU states (in 2002) plus Switzerland and Norway
** As defined by NGO network connectivity (see equation (3)).
Table 2: Top 25 NGO cities by network connectivities
Table 3: 7 Articulator Cities
Figure 1: Activity Field for Component I
(For city codes see Figure BI (appendix B)
Figure 2: Activity Field for Component II
(For city codes see Figure BI (appendix B)
Figure 3: Activity Field for Component III
(For city codes see Figure BI (appendix B)
Figure 4: Activity Field for Component IV
(For city codes see Figure BI (appendix B)
Figure 5: Activity Field for Component V
(For city codes see Figure BI (appendix B)
Figure 6: Activity Field for Component VI
(For city codes see Figure BI (appendix B)
Figure 7: Activity Field for Component VII
(For city codes see Figure BI (appendix B)
Figure 8: Activity Field for Component VIII
(For city codes see Figure BI (appendix B)
Figure 9: Activity Field for Component IX
(For city codes see Figure BI (appendix B)
Figure 10: Activity Field for Component X
(For city codes see Figure BI (appendix B)
APPENDIX A NON-GOVERNMENTAL ORGANIZATIONS
i 58 NGOs included in the principal components analysis
ii 16 additional NGOs included in computing connectivities
APPENDIX B CITIES
i 100 cities included in the principal components analysis
Figure B1: Cartogram of 100 NGO cities
(Cities are shown in their approximate geographical locations) City codes are as follows: AA Addis Ababa; AB Abidjan; AC Accra; AK Ankara; AL Algiers; AM Amsterdam; AN Amman; AO Antananarivo; AT Athens; AU Auckland; AY Almaty; BA Buenos Aires; BC Barcelona; BD Budapest; BG Bogota; BJ Beijing; BK Baku; BM Bamako; BN Bangkok; BR Brussels; BS Brasilia; BT Bridgetown; BU Bucharest; CA Cairo; CH Chennai; CO Colombo; CP Copenhagen; CR Caracas; CT Cotonou; DA Dakar; DB Dublin; DH Dhaka; DS Dar es Salam; FR Frankfurt; FT Freetown; GC Guatemala City; GN Geneva; GT Georgetown; HK Hong Kong; HL Helsinki; HN Hanoi; HR Harare; IS Islamabad; JB Johannesburg; JK Jakarta; KM Kathmandu; KN Kinshasa; KP Kampala; KS Kingston; KV Kiev; LA Los Angeles; LG Lagos; LM Lima; LN London; LO Lome; LP La Paz; LU Luanda; MD Madrid; ME Melbourne; MG Managua; MN Manila; MO Moscow; MP Maputo; MT Montreal; MU Mumbai; MV Montevideo; MX Mexico City; NA Nairobi; ND New Delhi; NY New York; OS Oslo; OT Ottawa; OU Ouagadougou; PA Paris; PH Phnom Penh; PR Pretoria; QU Quito; RB Rabat; RJ Rio de Janeiro; RO Rome; SA Santiago; SD Santa Domingo; SE Seoul; SG Singapore; SJ San Jose; SP Sao Paulo; SS San Salvador; ST Stockholm; SV Suva: SY Sydney; TE Tegucigalpa; TP Taipei; TK Tokyo; TR: Toronto; UL Ulaanbaatar; VI Vienna; WA Washington, DC; WS Warsaw; YA Yaounde; ZG Zagreb
ii 78 additional cities included in computing connectivities
APPENDIX C COMPONENT LOADINGS
Edited and posted on the web on 6th August 2003
Note: This Research Bulletin is a source paper for RB 144 and will not be published in its present form.